臺灣地勢陡峻、河川坡短流急，因此於颱洪時期水庫適當的操作，達到消災減洪並有效蓄水為重要之議題。在平行水庫聯合防洪操作時，需同時考量洪峰消減率與洪峰時間稽延，並避免洪峰同時抵達匯流口造成水患災情等問題；因此，本研究探討平行水庫於颱洪時期聯合防洪操作時，除了考慮水庫防洪與蓄水外，同時考慮匯流口之洪水歷程，以期有效地提升洪峰消減率之效。 本研究以石門、翡翠水庫系統及下游匯流口為研究對象，建立兩平行水庫最佳化防洪操作模式，利用非優勢排序遺傳演算法(NSGA-II)搜尋平行水庫防洪最佳放流歷程。NSGA-II搜尋結果與原始操作結果比較顯示，NSGA-II獲得較佳之水庫洪峰消減率；無論是水庫或匯流口洪峰消減率也僅少數幾場與原操作相近，其餘皆達較好之洪峰消減率。綜合結果得知NSGA-II搜尋之放流歷程可有效達到較佳之洪峰消減率並降低匯流口洪峰，最後滿足蓄水之目標以利未來水資源運用。 Taiwan is located in the southeast of Asia where always hit by Typhoon. Because of the steep mountainous landform, most of the rainfall flow immediately into the ocean within a few hours. Reservoirs operation have become the most important issue during Typhoon periods. In the parallel reservoirs operation, it is crucial to ponder both the decreasing rate and time lag of flood peak and the prevention of disaster caused by flood peak arriving at confluence at the same time. Therefore, this study investigates optimal parallel reservoir system flood operation in typhoon periods not only to balance between the reservoir flood avoidance and water storage improvement but also to decrease the downstream. This study takes the Shihmen and Feitsui reservoir system and the downstream confluence as research objects to build up the parallel reservoir system flood operation optimal model by adopting the non-dominate sorting genetic algorithm (NSGA-II) as search engine. Comparing to the traditional operation, NSGA-II gains better decreasing rate of flood peak both on reservoir system and downstream confluence even if a few performances are close. In conclusion, the results demonstrate that NSGA-II obtains better decreasing rate of flood peak, reduces the flood damage downstream and increases final storage for future usages.
